Claude Sonnet 4.6 benchmarks

Claude Sonnet 4.6 is a proprietary model from Anthropic, released 17 Feb 2026. It ranks 39th of 51 on our overall leaderboard and 24th of 51 for long context. At A$8.42 per million tokens (blended), it costs 2.3× the median model we track. It can run with requests kept in Australia on AWS Bedrock in Sydney and Melbourne.

#39 OF 51 OVERALLA$8.42 / 1M TOKENS1M CONTEXTUPDATED 23 SEP 2026
CONTEXT WINDOW
1,000,000 tokens
MAX OUTPUT
128,000 tokens
INPUT
text, image
WEIGHTS
Proprietary
RELEASED
17 Feb 2026
ANTHROPIC DOCS
§ 02 — RESULTS

Benchmark results.

Independent test scores are for the adaptive reasoning, max effort setting. “Among tracked” ranks Claude Sonnet 4.6 against the 51 models on this site; LMArena ranks run across its full leaderboard.

ALL RESULTS7 OF 8 CORE SIGNALS
Claude Sonnet 4.6 benchmark results
BENCHMARKRESULTAMONG TRACKEDSOURCE DETAIL
INDEPENDENT TESTS · ARTIFICIAL ANALYSIS
AA Intelligence Index30.1#40 of 51Adaptive reasoning, max effort
AA Coding Index63.0#33 of 42Adaptive reasoning, max effort
GPQA Diamond87.5%#39 of 44Adaptive reasoning, max effort
Humanity’s Last Exam33.6%#46 of 51Adaptive reasoning, max effort
AA-LCR80.0%#27 of 51Adaptive reasoning, max effort
Terminal-Bench 2.171.2%#33 of 42Adaptive reasoning, max effort
τ²-Bench (banking)34.4%#25 of 41Adaptive reasoning, max effort
SciCode50.1%#39 of 46Adaptive reasoning, max effort
BLIND HUMAN VOTES · LMARENA
LMArena Text66,208 votes · ±41473#26 of 42#37 of 402 on LMArena · 13 Sep 2026
LMArena WebDev21,659 votes · ±51521#29 of 45#36 of 129 on LMArena · 22 Sep 2026
LMArena Vision25,552 votes · ±61275#21 of 33#29 of 152 on LMArena · 13 Sep 2026
LMArena Document57,813 votes · ±61478#10 of 26#13 of 44 on LMArena · 13 Sep 2026
LMArena Search134,905 votes · ±51201#9 of 12#10 of 34 on LMArena · 24 Aug 2026
LMArena Agent39,452 sessions#31#27 of 37#31 of 46 on LMArena · 15 Sep 2026
§ 03 — COST

What Claude Sonnet 4.6 costs in Australian dollars.

List prices converted at the snapshot’s RBA rate, excluding GST. Reasoning models are billed for their thinking as output tokens, so heavier settings cost more than the job estimates show.

LIST PRICE
Claude Sonnet 4.6 price per million tokens
PER MILLION TOKENSAUDUSD LIST
Input tokensA$4.21US$3.00
Output tokensA$21.06US$15.00
Blended (3 in : 1 out)A$8.42US$6.00
EVERYDAY JOBSESTIMATES
Estimated cost of common jobs in Australian dollars
JOBCOST (AUD)MEDIAN MODEL
Customer support replyPER 1,000 REPLIESA$14.74A$6.58
Summarise a 30-page documentPER 100 DOCUMENTSA$10.11A$4.74
Agentic coding taskPER 10 TASKSA$8.00A$3.72
SETTINGSMEASURED SEPARATELY
Claude Sonnet 4.6 settings compared
SETTINGINTELLIGENCEA$ / 1MSPEED
Adaptive reasoning, max effort30.1A$8.42
Non-reasoning, high effort24.7A$8.42
Non-reasoning, low effort23.3A$8.42
§ 04 — AUSTRALIA

Running Claude Sonnet 4.6 in Australia.

It can run with requests kept in Australia on AWS Bedrock in Sydney and Melbourne. How the platforms compare →

AUSTRALIAN CLOUD REGIONS
Claude Sonnet 4.6 availability in Australian cloud regions
PLATFORMAVAILABILITYDETAIL
AWS Bedrock · SydneyAU ROUTING

Provider docs, checked 23 Sep 2026

AWS Bedrock · MelbourneAU ROUTING

Provider docs, checked 23 Sep 2026

Azure · Australia EastNOT IN AU

Provider docs, checked 23 Sep 2026

Google Vertex AI · SydneyGLOBAL

Provider docs, checked 23 Sep 2026

§ 06 — QUESTIONS

Claude Sonnet 4.6, answered.

Claude Sonnet 4.6 lists at US$3.00 per million input tokens and US$15.00 per million output tokens — A$4.21 and A$21.06 at A$1 = US$0.7123 (RBA, 22 Sep 2026), excluding GST. A typical customer support reply works out at about A$14.74 per 1,000 replies, before any reasoning tokens.

Claude Sonnet 4.6 can run with requests kept in Australia on AWS Bedrock in Sydney and Melbourne. Availability comes from the cloud providers’ own documentation; check your provider’s current terms before relying on it for data-residency obligations.

It ranks 33rd of 42 on our coding ranking, with an Artificial Analysis Coding Index of 63.0 and an LMArena WebDev rating of 1521 (36th of 129). The current leader is Claude Fable 5.1.

It ranks 30th of 41 on our agents ranking, completing 34.4% of τ²-Bench banking customer-service tasks and 71.2% of Terminal-Bench 2.1 tasks.

Anthropic documents a context window of 1,000,000 tokens (1M), with up to 128,000 output tokens. On AA-LCR, which tests reasoning across ~100,000-token document sets, it scores 80.0%.

PUT THE COMPARISON TO WORK

Need help choosing and using AI for your business?